Please note: This seminar will be given online.
Dallas
Card, Postdoctoral
scholar
NLP
Group
and
the
Data
Science
Institute, Stanford
University
Machine learning and natural language processing have become increasingly influential, both in commercial applications and as key tools for research in the natural and social sciences. In both cases, however, research in these fields raises numerous concerns related to bias, transparency, robustness, and how we communicate information.
In this talk, I will discuss two separate but related strands of work: one on developing more reliable approaches to machine learning, and the other on computational methods for learning about society from text. In the first part I will present work on modifying deep learning classifiers to allow for more transparent explanations, while also providing greater robustness to out of domain data. In the second, I will focus on using computational tools to analyze how different media sources present opinions about climate change, demonstrating bias in both news framing and annotator perceptions.
Bio: Dallas Card is a postdoctoral scholar in the NLP group and the Data Science Institute at Stanford University. His work focuses on making machine learning more reliable and responsible, both in research and application, and on computational social science using textual data. Dallas is a graduate of Systems Design Engineering at the University of Waterloo, and holds a PhD in machine learning from Carnegie Mellon University.
To join this seminar on Zoom, please go to https://zoom.us/j/96624131528?pwd=ZkwwYkNDdTNBT0YrY2w4RC9KSUtNQT09.